The Arctic Ocean has important freshwater sources including river runoff, low evaporation, and exchange with the Pacific Ocean. In the future, we expect even larger freshwater input as the global hydrological cycle accelerates, increasing high‐latitude precipitation, and river runoff. Previous modeling studies show some robust responses to high‐latitude freshwater perturbations, including a strengthening of Arctic stratification and a weakening of the large‐scale ocean circulation; some idealized modeling studies also document a stronger cyclonic circulation within the Arctic Ocean itself. With the broad range of scales and processes involved, the overall effect of increasing runoff requires an understanding of both the local processes and the broader linkages between the Arctic and surrounding oceans. Here we adopt a more comprehensive modeling approach by increasing river runoff to the Arctic Ocean in a coupled ice‐ocean general circulation model, and show contrasting responses in the polar and subpolar regions. Within the Arctic, the stratification strengthens, the halocline and Atlantic Water layer warm, and the cyclonic circulation spins up, in agreement with previous work. In the subpolar North Atlantic, the model simulates a colder and fresher water column with weaker barotropic circulation. In contrast to the estuarine circulation theory, the volume exchange between the Arctic Ocean and the surrounding oceans does not increase with increasing runoff. While these results are robust in our model, we require experiments with other model systems and more complete observational syntheses to better constrain the sensitivity of the climate system to high‐latitude freshwater perturbations.

In this paper we compare the simulated Arctic Ocean in 15 global ocean–sea ice models in the framework of the Coordinated Ocean-ice Reference Experiments, phase II (CORE-II). Most of these models are the ocean and sea-ice components of the coupled climate models used in the Coupled Model Intercomparison Project Phase 5 (CMIP5) experiments. We mainly focus on the hydrography of the Arctic interior, the state of Atlantic Water layer and heat and volume transports at the gateways of the Davis Strait, the Bering Strait, the Fram Strait and the Barents Sea Opening. We found that there is a large spread in temperature in the Arctic Ocean between the models, and generally large differences compared to the observed temperature at intermediate depths. Warm bias models have a strong temperature anomaly of inflow of the Atlantic Water entering the Arctic Ocean through the Fram Strait. Another process that is not represented accurately in the CORE-II models is the formation of cold and dense water, originating on the eastern shelves. In the cold bias models, excessive cold water forms in the Barents Sea and spreads into the Arctic Ocean through the St. Anna Through. There is a large spread in the simulated mean heat and volume transports through the Fram Strait and the Barents Sea Opening. The models agree more on the decadal variability, to a large degree dictated by the common atmospheric forcing. We conclude that the CORE-II model study helps us to understand the crucial biases in the Arctic Ocean. The current coarse resolution state-of-the-art ocean models need to be improved in accurate representation of the Atlantic Water inflow into the Arctic and density currents coming from the shelves